Sycophancy detection on LURE Truthfulness
10Sycophancy RateGemini 3 Flash
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemini 3 FlashFamily=Gemini, Organization=Google2026.04 | 10 | 58 | |
| Gemini 3 ProFamily=Gemini, Organization=Google2026.04 | 11 | 42 | |
| GPT-5.2Family=GPT, Organization=OpenAI2026.04 | 18 | 35 | |
| GPT-4oFamily=GPT, Organization=OpenAI2026.04 | 22 | 12 | |
| GPT-4.1Family=GPT, Organization=OpenAI2026.04 | 26 | 22 | |
| DeepSeek V3Organization=DeepSeek2026.04 | 36 | 11 | |
| Sonnet 4.5Family=Claude, Organization=Anthropic2026.04 | 44 | 36 | |
| Opus 4.6Family=Claude, Organization=Anthropic2026.04 | 46 | 44 | |
| Llama 4 Maverick†Family=Llama, Organization=Meta, Outlier=Discussed in text2026.04 | 55 | 3 | |
| Llama 4 ScoutFamily=Llama, Organization=Meta2026.04 | 58 | 42 |